AI-102 Practice Test 32 — 15 Questions

Question 1

You are implementing a generative AI solution using Azure OpenAI Service. The solution must generate responses that are grounded in your organization's proprietary documents and must return citations that link back to the source documents. You need to configure the deployment to meet these requirements. Which two actions should you perform? (Choose two.)

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Create an Azure AI Search index that contains the proprietary documents with retrievable content and citation fields such as title and URL.. Grounded responses with citations in Azure OpenAI require a retrieval source and a deployment-level data source binding. An Azure AI Search index holds the proprietary documents with content and citation metadata, and associating that index with the model deployment through On Your Data enables service-side retrieval, prompt injection, and citation generation.

Answer

Enable diagnostic logging on the Azure OpenAI resource to capture prompt and completion text.

Answer

Fine-tune the model on the proprietary documents so it can reproduce them verbatim.

Answer

Deploy a second model instance in a different Azure region for high availability.

Question 2

You are deploying a retrieval-augmented generation chat solution on Azure OpenAI. During testing, users report that the model confidently answers questions using facts that are not in the indexed documents, and that retrieved chunks sometimes come from the wrong department's SharePoint site. You need to reduce these ungrounded answers without retraining the model. (Choose two.)

Accepted Answer

Set the message role "system" content to instruct the model to answer only from the provided context and to state when the answer is not found.. Ungrounded answers in a RAG pattern usually come from two sources: the model not being constrained to the retrieved context, and the retrieval step returning irrelevant or unauthorized chunks. A system message that enforces answer-only-from-context behavior, combined with metadata filtering and security trimming at the search layer, addresses both causes without retraining. Length and randomness parameters, plus harmful-content filters, do not influence factual grounding.

Answer

Set the temperature parameter to 2 to make the model more factual.

Answer

Increase the max_tokens parameter so the model has more room to explain its reasoning before answering.

Answer

Enable content filtering at the highest severity threshold to block ungrounded statements.

Question 3

You are building a solution that uses Azure AI Language's question answering (custom question answering) to create a bot that answers employee HR questions. You have created a project and added a knowledge base with 50 question-answer pairs. You need to ensure the bot provides accurate answers and can handle paraphrased questions. Which two actions should you perform? (Choose two.)

Accepted Answer

Add alternative questions for each question-answer pair to cover different ways employees might ask the same question.. To improve paraphrase handling in custom question answering, you should add alternative questions to cover different phrasings and enable active learning to continuously refine the knowledge base based on user queries. Increasing thresholds or adding entities does not help with matching paraphrased questions.

Answer

Configure the project to use a different language model for better semantic understanding.

Answer

Increase the confidence threshold for answer matching to ensure only high-confidence answers are returned.

Answer

Add a prebuilt entity for 'person' to the project to extract employee names from questions.

Question 4

A travel agency wants its web app to summarize long customer complaint emails into a short paragraph that preserves the main points. The emails are in English and average 800 words. You need to use the extractive summarization feature of Azure AI Language and control how long the summary is. Which request parameter should you configure?

Accepted Answer

maxSentenceCount. Extractive summarization returns the highest-ranked sentences from the input, so length is controlled by limiting how many sentences are returned. maxSentenceCount sets that cap and directly satisfies the requirement to produce a shorter or longer summary while retaining the original phrasing of the complaint email. The other parameters either belong to different tasks or do not exist on this request.

Answer

summaryLength

Answer

sentenceCountToExclude

Answer

temperature

Question 5

You are building a generative AI feature in an application using the Azure OpenAI SDK. The feature must stream tokens to the user as they are generated and must also capture the full response for logging. Which implementation approach satisfies both requirements?

Accepted Answer

Set stream=true on the chat completions request, iterate over the streamed chunks to display deltas, and accumulate the delta content to form the complete response for logging.. Streaming chat completions deliver delta content across multiple chunk objects. Rendering those deltas satisfies the progressive display requirement, and concatenating them rebuilds the exact full message for logging. A single streaming request therefore meets both needs, whereas duplicate calls, partial logging, or multiple completions fail one requirement or the other.

Answer

Disable streaming and set n=2 so the model returns two completions, displaying one and logging the other.

Answer

Make two separate non-streaming requests: one to display the response and one to log it.

Answer

Set stream=true and log only the first chunk received from the stream.

Question 6

A legal team wants an assistant that drafts contract summaries. Their policy requires that every generated summary include traceable references to the exact clauses used and that reviewers be able to see which source passages informed each statement. You are using Azure OpenAI with your own document index. Which approach best meets the traceability requirement?

Accepted Answer

Use the On Your Data pattern with Azure AI Search, return document chunks with their IDs and titles, and instruct the model to cite the retrieved chunk identifiers in its output.. Traceability requires linking generated text back to specific source passages. Using Azure AI Search as the grounding source with the On Your Data pattern returns citation metadata for the retrieved chunks, and prompting the model to cite those chunks produces summaries where each statement can be traced to a clause. Quota, fine-tuning, and logprobs do not create that link.

Answer

Fine-tune the base model on previously approved contract summaries so its style matches legal expectations.

Answer

Enable the model's logprobs parameter and expose the token probabilities to reviewers as evidence of reliability.

Answer

Increase the model deployment's tokens-per-minute quota so longer contracts fit in a single prompt.

Question 7

You are building an internal knowledge assistant with Azure OpenAI Service. Responses must be grounded in a curated set of HR policy documents stored in Azure AI Search, and every response must include a citation to the specific source chunk. You already deployed a GPT-4o model and created the search index. You need to add the grounding layer with the least development effort. What should you do?

Accepted Answer

Enable the "On Your Data" feature on the model deployment and point it at the Azure AI Search index.. Grounding a deployment in a private index with automatic citations is exactly what the On Your Data capability provides for Azure OpenAI. Since the Azure AI Search index is already built, enabling this feature on the deployment satisfies the grounding and citation requirements with configuration instead of custom code, which matches the least-effort constraint.

Answer

Increase the model temperature and max tokens so the model has more room to recall the HR policies from its pretrained knowledge.

Answer

Call the Azure AI Search REST API from the client, concatenate the top results into the system message, and post-process the model output to append source links.

Answer

Fine-tune the GPT-4o deployment on the HR policy documents, then rely on the model to quote the correct policy section.

Question 8

You are prototyping a chat experience on Azure OpenAI and want the model to produce structured JSON that matches a schema your application can deserialize reliably. Which feature should you configure?

Accepted Answer

Set response_format to json_schema with a strict schema definition on the chat completions call.. Schema-constrained generation is the only option that enforces structure at decode time. Configuring response_format with a strict JSON schema makes the model emit payloads that conform to the defined fields and types, so the application can deserialize without defensive parsing. Sampling parameters, prompt hints, and content filters do not enforce schema conformance.

Answer

Enable content filtering and set the severity threshold to high.

Answer

Increase top_p to 1 and rely on the model's instruction-following ability.

Answer

Set temperature to 0 and add the word JSON to the user prompt.

Question 9

You are implementing a Retrieval Augmented Generation pattern with Azure AI Search and Azure OpenAI. Users complain that answers occasionally cite documents the user is not authorized to see, because the index contains all departments' content. You need to restrict retrieval so that each user only receives chunks from documents their identity permits. What should you do?

Accepted Answer

Add a security filter on a document access field in Azure AI Search, and pass the user's group or object IDs as an OData filter in the query.. Security trimming for RAG must occur at retrieval. Storing each document's allowed principals in a filterable field and passing the caller's group or object IDs as an OData filter ensures the search response never contains unauthorized chunks. Relevance ranking, prompt instructions, and manual index selection do not enforce authorization and can leak restricted content before the model or user sees it.

Answer

Create a separate index per department and have users manually select which index to search.

Answer

Add a system message instructing the model to ignore any content the user is not allowed to see.

Answer

Apply a semantic ranker to the query and rely on relevance scoring to push unauthorized content lower in the results.

Question 10

A company is building a computer vision solution using Azure AI Vision to analyze images of retail shelves. The solution must detect product presence and read expiration dates. Which TWO Azure AI Vision features should be used?

Accepted Answer

Object detection. Object detection (C) is correct because it locates and classifies multiple objects within an image, which is exactly what is needed to determine whether specific products are present on retail shelves. Optical Character Recognition (D) is correct because OCR extracts printed or handwritten text from images, enabling the solution to read expiration dates printed on product packaging. Face detection (A) only identifies human faces and their attributes, which is irrelevant to detecting products or reading dates. Brand detection (B) identifies known company logos but does not determine product presence or read expiration dates. Image captioning (E) generates a natural-language description of an image and cannot reliably detect specific products or extract date text.

Answer

Face detection

Answer

Brand detection

Answer

Image captioning

Question 11

You are building a generative AI feature that drafts marketing copy in English and must also produce accurate French and German versions. The team wants a single deployment that handles all three languages without maintaining separate prompt templates per language, and the translations must preserve the marketing tone. Which Azure OpenAI capability should the solution rely on?

Accepted Answer

Use a single model deployment and instruct it in the system message to generate in the requested language while preserving tone.. Azure OpenAI chat models are trained on multilingual data and respond well to explicit language and style instructions. Specifying the target language and the marketing tone in the system message lets one deployment serve English, French, and German without maintaining separate templates. Separate deployments, an external translation pipeline, or fine-tuning all add complexity that the scenario does not require and do not better guarantee tone preservation.

Answer

Translate the English output using Azure AI Translator and then post-process with a grammar model.

Answer

Deploy a separate model instance per language and route requests based on the target locale.

Answer

Fine-tune the model on parallel marketing copy in all three languages before deploying it.

Question 12

You are building a knowledge mining solution with Azure AI Search. The solution indexes scanned PDF reports stored in Azure Blob Storage. Each report contains multiple embedded images with text that must be searchable. You have already created a data source, index, and indexer. You need to ensure that text from the embedded images is extracted and mapped to the 'content' field in the index. Which two actions should you perform? (Choose two.)

Accepted Answer

Enable image extraction by setting the 'imageAction' parameter to 'generateNormalizedImages' in the indexer definition.. To make text from embedded images searchable, you must first extract the images from the documents by setting imageAction to generateNormalizedImages. Then, you add an OCR skill with context /document/normalized_images/* to process each image and output the recognized text. This text can then be mapped to the content field. Without both actions, the OCR skill either has no input or does not run per image.

Answer

Configure the indexer to use the 'text' parsing mode instead of the default 'json' mode.

Answer

Add an Entity Recognition skill to detect and extract text from images.

Answer

Add a Shaper skill that merges all image text into a single string before indexing.

Question 13

You are deploying a generative AI solution that must produce structured JSON order confirmations from free-text customer emails. Downstream systems reject any response that is not valid JSON matching a fixed schema. Which approach best guarantees schema-conformant output from an Azure OpenAI chat completion?

Accepted Answer

Use a structured outputs configuration with a strict JSON schema that defines required fields and types.. Structured outputs constrain token generation to a supplied JSON schema, enforcing required fields and correct types so responses validate against the contract. JSON mode only guarantees syntactic validity, and few-shot examples or post-processing repairs do not provide a hard structural guarantee. For downstream systems that reject any schema deviation, constrained decoding is the most reliable approach.

Answer

Set the temperature to zero and include an example of the exact JSON output in the prompt.

Answer

Set response_format to json_object and describe the desired schema in the system message.

Answer

Request the output in JSON and run a regex-based validator that repairs any deviations before forwarding.

Question 14

A company uses Azure Document Intelligence to process purchase orders. They have trained a custom model with 10 labeled samples and deployed it as 'purchaseOrderModel'. When analyzing a new purchase order, the extracted 'TotalAmount' field is often incorrect. The company wants to improve the model's accuracy for this field. What should they do?

Accepted Answer

Retrain the model with additional labeled samples that include variations of the 'TotalAmount' field.. Custom model accuracy in Azure Document Intelligence depends heavily on the quality and quantity of labeled training data. When a specific field like TotalAmount is frequently misidentified, the most effective action is to retrain with more labeled examples that capture the field's variations. This helps the model learn the patterns and contexts associated with that field, leading to better generalization on new documents.

Answer

Switch to the prebuilt invoice model, which automatically extracts total amounts from purchase orders.

Answer

Increase the model's confidence threshold for the 'TotalAmount' field in the project settings.

Answer

Add a labeled sample where the 'TotalAmount' field is left blank to teach the model to ignore missing values.

Question 15

You are designing an Azure AI Search enrichment pipeline that processes scanned PDF invoices stored in Azure Blob Storage. You need to extract both printed text and handwritten notes from the documents before sending the content to an Azure AI Language entity recognition skill. The solution must minimize development effort and cost. Which skill should you add to the skillset?

Accepted Answer

Microsoft.Skills.Vision.OcrSkill. The built-in OCR skill reads printed and handwritten text from images and PDFs and writes the extracted text into the enrichment tree. Because it is a first-party Azure AI Search skill, it requires no custom code, which satisfies the goals of low effort and predictable cost. Other text skills assume machine-readable input and cannot perform image recognition.

Answer

Microsoft.Skills.Text.MergeSkill

Answer

Microsoft.Skills.Custom.WebApiSkill

Answer

Microsoft.Skills.Text.KeyPhraseExtractionSkill